CN109033101B - Label recommendation method and device - Google Patents
Label recommendation method and device Download PDFInfo
- Publication number
- CN109033101B CN109033101B CN201710429233.7A CN201710429233A CN109033101B CN 109033101 B CN109033101 B CN 109033101B CN 201710429233 A CN201710429233 A CN 201710429233A CN 109033101 B CN109033101 B CN 109033101B
- Authority
- CN
- China
- Prior art keywords
- tag
- label
- entity
- target
- information
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
- 238000000034 method Methods 0.000 title claims abstract description 61
- 230000006399 behavior Effects 0.000 claims abstract description 40
- 239000013598 vector Substances 0.000 claims description 94
- 230000006870 function Effects 0.000 claims description 19
- 238000012545 processing Methods 0.000 claims description 18
- 238000001914 filtration Methods 0.000 claims description 4
- 230000008569 process Effects 0.000 description 16
- 238000010586 diagram Methods 0.000 description 14
- 238000006073 displacement reaction Methods 0.000 description 8
- 238000012549 training Methods 0.000 description 4
- 238000013459 approach Methods 0.000 description 2
- 238000004891 communication Methods 0.000 description 2
- 238000004590 computer program Methods 0.000 description 2
- 238000010276 construction Methods 0.000 description 2
- 238000013500 data storage Methods 0.000 description 2
- 230000000694 effects Effects 0.000 description 2
- 238000005516 engineering process Methods 0.000 description 2
- 230000003287 optical effect Effects 0.000 description 2
- 101100281516 Caenorhabditis elegans fox-1 gene Proteins 0.000 description 1
- 230000006978 adaptation Effects 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000007423 decrease Effects 0.000 description 1
- 230000003247 decreasing effect Effects 0.000 description 1
- 239000000835 fiber Substances 0.000 description 1
- 238000009472 formulation Methods 0.000 description 1
- 238000011478 gradient descent method Methods 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 239000000203 mixture Substances 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 230000000717 retained effect Effects 0.000 description 1
- 239000004065 semiconductor Substances 0.000 description 1
- 239000007787 solid Substances 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q30/00—Commerce
- G06Q30/02—Marketing; Price estimation or determination; Fundraising
- G06Q30/0201—Market modelling; Market analysis; Collecting market data
Landscapes
- Business, Economics & Management (AREA)
- Strategic Management (AREA)
- Engineering & Computer Science (AREA)
- Accounting & Taxation (AREA)
- Development Economics (AREA)
- Finance (AREA)
- Entrepreneurship & Innovation (AREA)
- Game Theory and Decision Science (AREA)
- Data Mining & Analysis (AREA)
- Economics (AREA)
- Marketing (AREA)
- Physics & Mathematics (AREA)
- General Business, Economics & Management (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
Abstract
The application discloses a tag recommendation method and device, and belongs to the field of big data. The method comprises the following steps: reference label information is obtained, the reference label information is used for indicating user behaviors generated by a user aiming at a target service, the reference label information comprises information of n reference labels, and n is larger than or equal to 2. And then determining target entities corresponding to each reference label in the n reference labels and the pre-acquired appointed label in the initial knowledge graph to obtain m target entities, wherein m is more than or equal to 2. And then, establishing association relations for the m target entities according to the user behaviors indicated by the reference label information and the designated labels to obtain a reference knowledge graph. And finally, determining a recommended label based on the obtained reference knowledge graph. The method and the device solve the problems that recommendation basis is single when labels are recommended to service personnel in the related art, and the recommended labels are single, enrich the label recommendation basis, and the recommended labels are more abundant and diversified and are used for determining the target user group.
Description
Technical Field
The application relates to the field of big data, in particular to a tag recommendation method and device.
Background
In some service industries, in order to provide appropriate services to specific users, service personnel need to determine a target user group and then provide appropriate services to the determined target user group. When determining a target user group, a tag is usually used to determine a user grouping rule, and then related user data is obtained based on the user grouping rule, and then the target user group is obtained based on the related user data. Wherein the label is used for indicating the service use condition of the user.
For example, in order to retain old users and prolong the online time of the old users, the old users are usually determined as a target user group based on the above procedure, and then contract renewal work or package adaptation work is performed on the target user group. However, currently, when recommending a service to a target user group, a service person only uses an experience tag (the experience tag is obtained by experience) to determine a user grouping rule, and in this case, if no reference is made to a historical recommendation record about the service, the service recommendation cannot be completed well, which is called a cold start problem. To solve this cold start problem, business personnel need to be recommended diversified labels.
In the related art, a recommendation device is generally used to recommend a tag, and the recommendation device generally recommends a tag based on similarity between a tag vector of an experience tag corresponding to a user and a tag vector of a tag to be recommended. Specifically, the recommendation device determines a tag vector of an experience tag based on a pre-constructed knowledge spectrogram, wherein the experience tag is determined according to the condition that the user uses the current service, and then determines the tag vector of the tag to be recommended based on the knowledge spectrogram. And then, judging whether the similarity of the two label vectors is greater than a preset value. And when the similarity of the two label vectors is greater than a preset value, recommending the label to be recommended to a service person. The knowledge graph is a network, the network comprises a plurality of nodes and edges representing semantic relations among the nodes, each node is an entity (entity), and the entities correspond to the tags.
However, the above process only adopts the knowledge graph with fixed content to recommend labels to service personnel, the recommendation basis is single, and the recommended labels are single.
Disclosure of Invention
In order to solve the problems that recommendation basis is single when labels are recommended to service personnel and recommended labels are single in the related art, the embodiment of the invention provides a label recommendation method and a label recommendation device. The technical scheme is as follows:
in a first aspect, a tag recommendation method is provided, where the method includes: reference label information is obtained, the reference label information is used for indicating user behaviors generated by a user aiming at a target service, the reference label information comprises information of n reference labels, and n is larger than or equal to 2. Then, determining a target entity corresponding to each reference label in the n reference labels and a pre-acquired appointed label in the initial knowledge graph to obtain m target entities, wherein m is larger than or equal to 2, and the appointed label is acquired based on a preset user clustering rule. And then establishing association relations for the m target entities according to the user behaviors indicated by the reference label information and the designated labels to obtain a reference knowledge graph, wherein the reference knowledge graph is used for reflecting the association relations of the m target entities. Finally, a recommended label is determined based on the reference knowledge-graph.
The method improves the knowledge graph by combining with the user behavior, and recommends the labels for the service personnel based on the improved knowledge graph, so that the content of the knowledge graph is not fixed any more, the label recommendation basis is enriched, and the recommended labels are more abundant and diversified.
By way of example, the target business may be a business in the automotive industry, transportation industry, operator industry, internet industry, and the like.
Optionally, the reference tag information includes co-occurrence relationship between reference tags obtained by user behavior.
In the embodiment of the invention, a plurality of reference labels with co-occurrence relation can be obtained based on the browsing behavior of the user. Multiple reference tags with co-occurrence relationships may also be derived based on the user's purchasing behavior. The embodiment of the invention does not limit the form of the user behavior.
Optionally, before obtaining the reference tag information, the method may further include: the designated label is obtained based on a predetermined user grouping rule, which is determined based on the target service.
Determining a recommendation label based on the reference knowledge-graph may include: a first vector of a first target entity and a second vector of an incidence relation related to the first target entity in a reference knowledge graph are determined, wherein the first target entity is any one of m target entities. And predicting an entity having a co-occurrence relationship with the first target entity by adopting a scoring function according to the first vector and the second vector to obtain expected entity information, wherein the expected entity information comprises the first target entity, the entity having the co-occurrence relationship with the first target entity and a scoring result corresponding to the first target entity. And then, acquiring expected label information corresponding to the specified label. Then, the tags corresponding to the target entities in the expected entity information are inquired from the expected tag information, and a first tag list is obtained based on the expected entity information. And finally, performing data redundancy processing on the first tag list to obtain a recommendation list, wherein the recommendation list comprises recommendation tags.
The first tag list may contain redundant data that refers to redundant predicted co-occurrence tags. In this case, data redundancy processing may be performed on the first tag list. That is, redundant data (i.e., redundant predicted tags with co-occurrence relationships) in the first tag list is processed, that is, data (i.e., predicted tags with co-occurrence relationships) in the first tag list is subjected to redundancy removal processing, so that service personnel can quickly obtain recommended tags according to the recommended list, and the formulation efficiency of the user grouping rule is improved.
Optionally, obtaining the designated tag based on a predetermined user clustering rule may include: the user clustering rules are stored in a text form, and then the appointed tags are extracted from the user clustering rules in the text form.
Optionally, the first tag list includes a sorting value corresponding to a tag, and performing data redundancy processing on the first tag list to obtain the recommendation list may include: the method comprises the steps of firstly, removing the repeated same labels in a first label list and reserving one label, and then taking the sum of the ranking values corresponding to each label in the same labels as the ranking value of the reserved corresponding label.
Optionally, the first tag list includes a sorting value corresponding to a tag, and data redundancy processing is performed on the first tag list to obtain a recommendation list, including: and removing the repeated same labels in the first label list and reserving one label, then carrying out weighted summation on the ranking value corresponding to each label in the same labels, and taking the result of the weighted summation as the ranking value of the reserved corresponding label.
Optionally, obtaining the designated tag based on a predetermined user clustering rule includes: the designated tag is periodically obtained based on a predetermined user clustering rule. Acquiring reference tag information, including: the reference tag information is periodically acquired.
In the embodiment of the invention, the user grouping operation can be fully utilized, and the user grouping rule formulated in the user grouping operation process is reused, so that the aim of updating the reference knowledge graph is fulfilled. Because the user clustering rule is formulated based on the experience label and the recommendation label, the label is obtained based on the user clustering rule, then the entity corresponding to the label and the reference label information for indicating the user behavior in the reference knowledge graph is determined, then the association relationship is established for the corresponding entity in the reference knowledge graph according to the user behavior indicated by the reference label information and the label obtained based on the user clustering rule, the reference knowledge graph can be further updated and optimized, and finally, the recommendation label determined based on the reference knowledge graph is continuously updated and optimized. The continuous updating and optimizing process improves the efficiency and the accuracy of business personnel for formulating the user grouping rule and improves the efficiency and the accuracy of determining the target user group.
Optionally, determining a target entity corresponding to each reference tag of the n reference tags and a pre-obtained designated tag in the initial knowledge graph to obtain m target entities, including: filtering preset entities in the initial knowledge graph by adopting a word frequency TF reverse file frequency IDF statistical mode; m target entities are determined from the filtered entities.
In the embodiment of the invention, in order to improve the construction efficiency of the initial knowledge graph, entities with lower importance degree in the initial knowledge graph can be filtered in advance, and then the reference label and the target entity corresponding to the specified label are determined from the rest entities.
Optionally, the expected tag information includes a tag name of the specified tag, a determination manner of the user clustering rule, a tag statistical period, and a tag entity of the specified tag.
Optionally, determining a first vector of the first target entity and a second vector of the association relation related to the first target entity in the reference knowledge-graph includes: and determining the first vector and the second vector by adopting a knowledge base mode.
In a second aspect, a tag recommendation apparatus is provided, which includes at least one module, where the at least one module is configured to implement the tag recommendation method according to the first aspect.
In a third aspect, a tag recommendation apparatus is provided that includes a processor, a memory, a network interface, and a bus. Wherein the bus is used for connecting the processor, the memory and the network interface. The memory may comprise random access memory or may comprise non-volatile memory, such as at least one disk memory. The communication connection between the recommendation device and the external device is realized through a network interface (which can be wired or wireless). The memory stores a program for implementing various application functions, and the processor executes the program stored in the memory to implement the tag recommendation method according to the first aspect.
In a fourth aspect, a computer-readable storage medium is provided, in which instructions are stored, and when the computer-readable storage medium runs on a computer, the computer is caused to execute the tag recommendation method of the first aspect.
In a fifth aspect, there is provided a computer program product comprising instructions which, when run on a computer, cause the computer to perform the tag recommendation method of the first aspect.
The technical effects obtained by the above second to fifth aspects are similar to the technical effects obtained by the corresponding technical means in the first aspect, and are not described herein again.
The technical scheme provided by the embodiment of the invention has the following beneficial effects:
reference label information can be obtained, then target entities corresponding to each reference label in n (n is larger than or equal to 2) reference labels and a pre-obtained designated label in an initial knowledge graph are determined, m (m is larger than or equal to 2) target entities are obtained, then association relations are established for the m target entities according to user behaviors and the designated labels indicated by the reference label information, a reference knowledge graph is obtained, and then a recommendation label is determined based on the reference knowledge graph. The reference label information is used for indicating user behaviors generated by a user aiming at target services, the reference label information comprises information of n reference labels, the designated labels are obtained based on a preset user clustering rule, and the reference knowledge graph is used for reflecting the incidence relation of m target entities. The knowledge graph is improved by combining with user behaviors, the labels are recommended for service personnel based on the improved knowledge graph, compared with the related technology, the content of the knowledge graph is not fixed any more, the label recommendation basis is enriched, and the recommended labels are more enriched and diversified.
Drawings
FIG. 1 is a schematic diagram of an implementation environment related to a tag recommendation method provided in some embodiments of the present disclosure;
fig. 2 is a schematic structural diagram of a tag recommendation device according to an embodiment of the present invention;
FIG. 3-1 is a flowchart of a tag recommendation method according to an embodiment of the present invention;
fig. 3-2 is a flowchart of acquiring a specific tag according to an embodiment of the present invention;
3-3 are schematic diagrams of an initial knowledge-graph;
3-4 are flow diagrams of a method for determining a target entity according to an embodiment of the present invention;
FIGS. 3-5 are schematic diagrams of a reference knowledge-graph provided by embodiments of the present invention;
3-6 are flow diagrams of determining recommended tags according to embodiments of the present invention;
3-7 are flow diagrams of a get recommendation list provided by embodiments of the present invention;
3-8 are flow diagrams of alternative lists for obtaining recommendations provided by embodiments of the present invention;
FIG. 4-1 is a schematic structural diagram of a tag recommendation apparatus according to an embodiment of the present invention;
FIG. 4-2 is a schematic structural diagram of another tag recommendation device provided in an embodiment of the present invention;
fig. 4-3 is a schematic structural diagram of a second determining module according to an embodiment of the present invention.
Detailed Description
To make the objects, technical solutions and advantages of the present application more clear, embodiments of the present application will be described in further detail below with reference to the accompanying drawings.
Fig. 1 is a schematic diagram of an implementation environment related to a tag recommendation method provided in some embodiments of the present disclosure, where the implementation environment may include: recommendation device 01 and external device 02.
The recommendation device 01 may be a server or a computer, and the recommendation device 01 is used for recommending tags for the service personnel.
The external device 02 may be a server, a server cluster composed of several servers, or a cloud computing service center. The recommendation device 01 is able to obtain data of the initial knowledge-graph from the external device 02. The recommendation device 01 and the external device 02 may establish a connection through a wired network or a wireless network.
The service personnel formulate a user grouping rule according to the label recommended by the recommendation equipment 01, then obtain related user data based on the formulated user grouping rule through the recommendation equipment 01, and obtain a target user group based on the related user data.
Fig. 2 is a schematic structural diagram of a tag recommendation apparatus according to an embodiment of the present invention, which may be used in the recommendation device 01 shown in fig. 1. As shown in fig. 2, the apparatus includes a processor 201 (e.g., a Central Processing Unit (CPU)), a memory 202, a network interface 203, and a bus 204. The bus 204 is used for connecting the processor 201, the memory 202 and the network interface 203. The Memory 202 may include a Random Access Memory (RAM) or a non-volatile Memory (non-volatile Memory), such as at least one disk Memory. The communication connection between the recommendation device 01 and the external device 02 is realized through a network interface 203 (which may be wired or wireless). The memory 202 stores therein a program 2021, the program 2021 is used to implement various application functions, and the processor 201 is used to execute the program 2021 stored in the memory 202 to implement the tag recommendation method described below.
Fig. 3-1 is a flowchart illustrating a tag recommendation method according to an exemplary embodiment, where the tag recommendation method is illustrated as being applied to the recommendation device 01 in the implementation environment shown in fig. 1, and the tag recommendation method may include:
And the service personnel formulates a user grouping rule based on the target service, and the recommendation equipment acquires the designated label based on the user grouping rule. By way of example, the target business may be a business in the automotive industry, transportation industry, operator industry, internet industry, and the like. The embodiment of the present invention is described by taking an example in which the target service is a service in the automobile industry.
As shown in fig. 3-2, step 301 may include:
For example, in order for a service person to provide a target service to a user who is ready to purchase a car, such as enabling the user to obtain reference information of a certain brand of car, and helping the user to make a car ordering plan, the service person may make a user grouping rule by using an experience tag. And then, the recommendation equipment stores the user grouping rule formulated by the service personnel in a text form. For example, the user grouping rule may be: the previous month prefers fox 1, the previous month prefers sedan 1, and the previous month prefers ford 1. Where 1 is used to indicate preference. For example, the previous month preference fox is 1, which indicates that the previous month preference fox is. And the previous month prefers fox to 0, which means that the previous month does not prefer fox.
Illustratively, the recommendation device derives the user grouping rule in text form from step 3011: in the "last month preference fox is 1, last month preference car is 1, and last month preference ford is 1", a specific tag is extracted, and the specific tag is: "prefer Fox", "prefer Car" and "prefer Ford".
When a user uses a target service, multiple tags are often used simultaneously, and in this case, the multiple tags are considered to have a co-occurrence relationship.
Optionally, the reference tag information may include co-occurrence relationships between reference tags resulting from user behavior.
For example, the reference tag information obtained by the recommendation device is: fox, 1.0T-1.6T, Cruz, 1.0T-1.6T. Where 1.0T represents a 1.0 liter capacity with turbocharging. In Fox and 1.0T-1.6T, a reference label Fox and a reference label 1.0T-1.6T have a co-occurrence relationship, and in Cruz and 1.0T-1.6T, a reference label Cruz and a reference label 1.0T-1.6T have a co-occurrence relationship.
For example, in the embodiment of the present invention, a plurality of reference tags having a co-occurrence relationship may be obtained based on the browsing behavior of the user. Specifically, for example, in the process of searching for information related to fox on the internet, the user inputs the keyword "fox" and also inputs the keyword "1.0T to 1.6T". That is, the user wants to search for information of Focus with a volume of 1.0T to 1.6T. In addition, for example, in the process of searching the information related to the crutz on the internet, the user inputs the keyword "crutz" and also inputs the keyword "1.0T to 1.6T", that is, the user wants to search the information of the crutz with the displacement of 1.0T to 1.6T.
Besides browsing behavior, a plurality of reference tags with co-occurrence relationship can be obtained based on the purchasing behavior of the user. Specifically, if the user buys fox, the displacement of the fox is also required to be 1.0T-1.6T; as another example, when a user buys Kouze, the displacement is also required to be 1.0T-1.6T.
The embodiment of the invention does not limit the form of the user behavior.
The recommendation device may obtain data of an initial knowledge map of the automotive industry from the external device. For example, part of the data of the initial knowledge-map obtained by the recommendation device from the external device may be as shown in table 1, where the name of the car, the train, the model, the brand, and the displacement are included in table 1. In addition, price and the like may also be included.
The recommendation device constructs an initial knowledge-graph from the data shown in table 1, and the constructed initial knowledge-graph may be as shown in fig. 3-3. In fig. 3-3, the entities having a connection relationship correspond to the first row data in table 1, the entities having a connection relationship correspond to the second row data in table 1, the entities having a connection relationship correspond to the third row data in table 1, and the entities having a connection relationship correspond to the fourth row data in table 1. 1.0L in table 1 indicates that the naturally aspirated engine has a displacement of 1.0 liter. The SUV in FIGS. 3-3 refers to a sport utility vehicle, which is in English: sports Utility Vehicle.
The recommending device constructs an initial knowledge graph, and then determines a target entity corresponding to the reference tag included in the reference tag information in step 302 and a target entity corresponding to the specified tag acquired in step 301 in the initial knowledge graph. Illustratively, the reference tag information obtained by the recommendation device in step 302 is: fox, 1.0T-1.6T, Cruz, 1.0T-1.6T. The recommendation device determines the target entities in the initial knowledge-graph shown in fig. 3-3 that correspond to the reference label "fox" in "fox, 1.0T-1.6T": fox, the target entity corresponding to the reference label "1.0T-1.6T" in Fox, 1.0T-1.6T ": 1.0T-1.6T; target entities corresponding to the reference label "collzis" in "collzis, 1.0T-1.6T": cruz, the target entity corresponding to the reference label "1.0T-1.6T" in "Cruz, 1.0T-1.6T": 1.0T to 1.6T. The recommendation device determines the target entity in the initial knowledge-graph shown in fig. 3-3 that corresponds to the designation label "preferred fox" in step 301: fox, the target entity corresponding to the designated tag "preferred cars": car, target entity corresponding to the specified label "preference ford": ford. Thus, the target entities determined by the recommendation device are: fox, 1.0T-1.6T, Cruz, Sedan and Ford.
TABLE 1
Further, in order to improve the construction efficiency of the initial knowledge graph, entities with lower importance degrees in the initial knowledge graph can be filtered in advance, and then the reference tags and the target entities corresponding to the designated tags are determined from the remaining entities. Optionally, as shown in fig. 3-4, step 303 may include:
And the recommending equipment filters preset entities in the initial knowledge graph in a Term Frequency (TF) reverse file Frequency (IDF) statistical mode.
The TF IDF statistical approach is used to evaluate the importance of a word (or phrase) to one of the documents in a corpus or a set of documents. The importance of a word (or phrase) increases in proportion to the number of times it appears in a document, but at the same time decreases in inverse proportion to the frequency with which it appears in the corpus. By way of example, the entity "car" with lower importance in the initial knowledge-graph shown in fig. 3-3 can be filtered out by TF IDF statistical means. The process of step 3031 may refer to the related art.
Since, in step 3031, the recommending apparatus filters out the entity "car" in the initial knowledge-graph, the recommending apparatus determines that the target entities corresponding to the reference tag and the designated tag are: fox, 1.0T-1.6T, Cruz, and Ford.
And step 304, the recommendation device establishes association relations for the m target entities according to the user behaviors indicated by the reference label information and the designated labels to obtain a reference knowledge graph, wherein the reference knowledge graph is used for reflecting the association relations of the m target entities.
And the recommendation equipment establishes association relations for the plurality of target entities obtained in the step 303 according to the user behaviors indicated by the reference label information in the step 302 and the specified labels in the step 301 to obtain the reference knowledge graph. For example, the reference tag information is: "Fox, 1.0T-1.6T", "Cruz, 1.0T-1.6T", the assigned labels are: "good fox", "good car", and "good ford", the target entities are: fox, 1.0T-1.6T, Cruz, and Ford. The recommendation device establishes associations for 4 target entities resulting in a reference knowledge-graph, which may be shown, for example, in fig. 3-5.
The dotted lines in FIGS. 3-5 indicate co-occurrence relationships, e.g., a co-occurrence relationship between Ford and Fox, and a co-occurrence relationship between Fox and (1.0T 1.6T). From FIGS. 3-5, a plurality of groups of co-occurring entities can be derived, such as Ford and Fox, Fox and (1.0T 1.6T), Ford and (1.0T 1.6T), Fox and Cruz. Entities having a vehicle relationship can also be derived from fig. 3-5 and the designated tags: 1 and Fox, 1 indicates a preference for Fox. Entities that have a displacement relationship can also be found: 1 and (1.0T to 1.6T) (i.e., preference (1.0T to 1.6T)).
In the embodiment of the invention, the recommended labels recommended to the service personnel can be determined based on the reference knowledge graph. Compared with the related technology, the reference knowledge graph integrates the user behaviors, the content of the knowledge graph is not fixed any more, and the recommendation labels are determined based on the reference knowledge graph and combined with the user behaviors, so that the label recommendation basis is richer, and the recommendation labels are richer and more diversified.
Optionally, as shown in fig. 3-6, step 305 may include:
Reference knowledge-maps as shown in fig. 3-5 will now be described. From fig. 3-5, a plurality of groups of entities having co-occurrence relationships can be obtained, including: ford and Fox, Fox and (1.0T-1.6T), Ford and (1.0T-1.6T), Fox and Cruz, entities that are in a vehicle relationship can also be obtained: 1 and Fox (i.e., preference Fox), entities that have a displacement relationship can also be derived: 1 and (1.0T to 1.6T) (i.e., preference (1.0T to 1.6T)). The 4 target entities are: fox, 1.0T-1.6T, Cruz, and Ford. The recommendation device may determine a first vector of "Focus," determine a second vector of associations (i.e., co-occurrence and train relationships) associated with "Focus"; determining a first vector of 1.0T-1.6T and a second vector of incidence relation (namely co-occurrence relation and displacement relation) related to 1.0T-1.6T; determining a first vector of 'Cruze', determining a second vector of incidence relations (i.e. co-occurrence relations) related to 'Cruze'; a first vector of "ford" is determined, and a second vector of associations (i.e., co-occurrence) related to "ford" is determined.
Specifically, the first target entity and the association relation related to the first target entity can be represented as a vector in a vector space by optimizing an interval-based loss function. For example, the first vector and the second vector may be obtained by using a knowledge base approach (transit), and the process may specifically refer to the related art.
And the recommending device predicts the entity with the co-occurrence relation with the first target entity by adopting a scoring function according to the first vector of the target entity and the second vector of the incidence relation related to the target entity. For example, the first vector of the target entity "Fox" is
The second vector of the association relation related to the target entity Fox is
A first vector
And a second vector
Adding to obtain a first vector
And a second vector
The sum vector of (1). Similarly, a sum vector of the first vector and the second vector of the target entity "1.0T-1.6T" may be obtained, a sum vector of the first vector and the second vector of the target entity "crutz" may be obtained, and a sum vector of the first vector and the second vector of the target entity "ford" may be obtained. And then, predicting the entity with the co-occurrence relation with the target entity by adopting a scoring function based on the 4 sum vectors to obtain expected entity information. The expected entity information comprises a target entity, an entity which is predicted to have a co-occurrence relation with the target entity and a scoring result corresponding to the target entity. For example, the expected entity information may be shown in table 2, where the ranking value in table 2 is the scoring result corresponding to the target entity, and the ranking value is also called a score.
TABLE 2
| Target entity | Predicted co-occurrence entities | Rank order value |
| Ford | Frelai snow | 0.8 |
| Ford | Modern day | 0.7 |
| Fox | Car (R.C.) | 0.3 |
| Fox | 1.0L～1.6L | 0.7 |
| Fox | Ix25 | 0.5 |
| Fox | Modern day | 0.5 |
Specifically, the recommendation device may generate a candidate entity set for a given target entity, score the candidate entities according to the entity relationship, and take the candidate entity with the highest score as the entity aligned with the target entity. The process may refer to the related art. And the ranking value of the candidate entity is also the ranking value of the target entity. For example, in table 2, the rank value of the candidate entity "snowfall" is 0.8, and the rank value of the target entity "ford" is also 0.8. "Freund" is an entity aligned with "Ford", and there is a co-occurrence relationship between "Freund" and "Ford".
In addition, when step 3051 is executed, the process of determining the vector of the entity and the vector of the association relationship may also be:
1) obtaining a positive case entity relationship pair set delta, a negative case entity relationship pair set delta' and a positive case set P formed by a relationship (namely an incidence relationship) r with a head entity h according to at least one entity relationship pair (h, r, t) of a knowledge graph (namely a reference knowledge graph)rT | (h, r, t) ∈ Δ } and a negative set of instances formed by the relation r with the head entity h
Wherein R represents a relationship set, an entity relationship pair (h, R, t)) The set delta' represents the set of entity relationship pairs (h, r, t) which exist in the knowledge-graph.
2) Initializing head entity vectors, relationship vectors and tail entity vectors in an entity relationship pair (h, r, t) of the knowledge graph according to given dimensions, wherein each head entity h corresponds to one head entity vector, each relationship r corresponds to one relationship vector, and each tail entity t corresponds to one tail entity vector.
3) According to the positive case set P for a specific entity h and corresponding relation rrAnd negative case set NrCalculating the entity interval M of a specific entity hh. Specifically, for a specific entity h and its corresponding relationship r, the selection is made
And
computing entity intervals Mh＝mint,t”δ (| | h-t "| - | | h-t |), wherein,
| | represents L1Or L2Normal form, mint,t”Means taking the minimum value from all the results calculated from t or t ".
4) According to the positive example entity relation pair set delta, the negative example entity relation pair set delta' and the entity interval MhA loss function is calculated.
The loss function is:
wherein M ishIndicates the entity interval corresponding to the head entity h, [ x ]]+Returning the greater of x and 0, | | | · | | represents L1Or L2A paradigm.
5) And iteratively updating a head entity vector, a relation vector and a tail entity vector of the entity relation pair, and when the loss function meets a preset condition, using the updated head entity vector, relation vector and tail entity vector as training models.
The iteratively updating the head entity vector, the relationship vector, and the tail entity vector of the entity-relationship pair may include:
updating by adopting a gradient descent method:
where dim is the dimension of the vector space, hiAn i-th dimension vector representing the h-vector of the head entity, mu is the learning rate,
hi＝hi-μ*2*|ti-hi-ri|，
ri＝ri-μ*2*|ti-hi-ri|，
ti＝ti+μ*2*|ti-hi-ri|，
h'i＝h'i-μ*2*|t'i-h'i-r'i|，
r'i＝r'i-μ*2*|t'i-h'i-r'i|，
t'i＝t'i-μ*2*|t'i-h'i-r'i|。
when step 3052 is executed, the process of scoring the entity according to the vector of the entity and the vector of the association relationship may also be:
6) reading the training model in the step 5), wherein the training model comprises the vector of the entity and the vector of the relationship.
7) And aiming at the given entity and the corresponding relation, constructing a candidate entity relation pair set according to the training model. The entity relationship pair set comprises at least one candidate entity relationship pair, each candidate entity relationship pair comprises a given entity, a relationship and a candidate entity, and the candidate entity is the same as the given entity in type.
8) And scoring the vectors of the entities in all the candidate entity relationship pairs and the vectors of the relationships according to a scoring function, and taking the candidate entity in the candidate entity relationship pair with the highest scoring value as an aligned entity, wherein the scoring function comprises the attribute similarity between the vector of the given entity and the vector of the candidate entity. The scoring function scores higher as the similarity value is higher.
Illustratively, when the given entity is the tail entity t, the corresponding relationship is r, and the candidate entity is the head entity h', the scoring function is
Wherein,
the vector similarity of h ' and t is represented, Dist (h ', t) represents the attribute similarity of h ' and t, and w represents the penalty degree, wherein the value range is 0 to 1.
Wherein Dist (h', t) ═ tt-h't|+EditDist(tattribute,h'attribute)，ttDenotes time of t, h'tTime, t, representing hattributeAttribute of t, h'attributeAn attribute representing h', EditDist (t)attribute,h'attribute) Indicating the edit distance between the attributes.
And 3053, acquiring expected tag information corresponding to the specified tag.
The recommended label acquires the expected label information from the external device. And data of user behaviors generated by the user aiming at the target service are stored in the external equipment.
With the assigned tag in step 301: "prefer Fox", "prefer Car", and "prefer Ford" are examples for illustration. The recommendation device obtains expected tag information corresponding to the specified tag, for example, the expected tag information may be as shown in table 3. The expected label information comprises a label name of a specified label, a determination mode of a user grouping rule, a label statistical period and a label entity of the specified label. For example, a tag name specifying the tag "prefer Fox" is: the preference Fox is that the determination mode of the user grouping rule corresponding to the specified label is as follows: the sum of the browsing times of the automobile home and the pacific automobile is more than 5, that is, when the sum of the browsing times of the automobile home and the pacific automobile is more than 5, the user is considered to prefer Focus. The tag statistical period of the designated tag is as follows: and carrying out monthly statistics and updating monthly. The tag entity of the designated tag is: fox, and Fox is a certain train.
TABLE 3
For example, the recommendation device queries the tags corresponding to the target entities "ford" and "fox" in table 2, i.e., "preferred fox" and "preferred ford", from the expected tag information shown in table 3, and obtains a first tag list based on table 2, which may be shown in table 4.
TABLE 4
| Label (R) | Predicted co-occurrence label | Rank order value |
| Preference Ford | Preference for Freilai | 0.8 |
| Preference Ford | Preference of modern times | 0.7 |
| Preference Fox | Preference car | 0.3 |
| Preference Fox | Preference is given to 1.0L to 1.6L | 0.7 |
| Preference Fox | Preference Ix25 | 0.5 |
| Preference Fox | Preference of modern times | 0.5 |
As shown in table 4, the first tag list may contain redundant data referring to redundant predicted co-occurrence tags. In this case, data redundancy processing may be performed on the first tag list. That is, redundant data (i.e., redundant predicted co-occurrence tags) in the first tag list is processed, i.e., data (i.e., predicted co-occurrence tags) in the first tag list is de-redundant. Referring to table 4, the first tag list may include sorting values corresponding to tags.
In one aspect, as shown in fig. 3-7, step 3055 can comprise:
step 3055a, one of the multiple identical tags repeatedly appearing in the first tag list is deduplicated.
For example, if 2 identical tags "prefer modern" are repeated among the tags predicted to have co-occurrence in table 4, then the recommendation device may "prefer modern" to duplicate one of the 2 identical tags. Then, the recommendation device takes the sum of the ranking values corresponding to the 2 same labels "preferred modern" as the ranking value of the label "preferred modern", that is, takes the sum of 0.5 and 0.7 as the ranking value of the label "preferred modern". The resulting recommendation list is shown in table 5, which includes the recommendation labels "prefer snowfall", "prefer modern", "prefer car", "prefer 1.0L-1.6L" and "prefer Ix 25".
TABLE 5
On the other hand, as shown in fig. 3-8, step 3055 can include:
step 3055c, one of the multiple identical tags repeatedly appearing in the first tag list is deduplicated.
And step 3055d, performing weighted summation on the ranking value corresponding to each label in the multiple same labels, and taking the result of the weighted summation as the retained ranking value corresponding to the label.
Also, taking table 4 as an example, the recommendation device may "prefer modern" to duplicate one over 2 identical tags. Then, the recommending device performs weighted summation on the ranking value 0.5 corresponding to one "preferred modern" and the ranking value 0.7 corresponding to the other "preferred modern", for example, the expression of weighted summation may be: 0.5 × 0.6+0.7 ═ 1-0.6 ═ 0.58. The recommendation device then takes 0.58 as the rank value of the label "prefer modern". The resulting recommendation list is shown in table 6, which includes the recommendation labels "prefer snowfall", "prefer modern", "prefer car", "prefer 1.0L to 1.6L", and "prefer Ix 25".
TABLE 6
| Recommendation label | Rank order value |
| Preference for Freilai | 0.8 |
| Preference of modern times | 0.58 |
| Preference car | 0.3 |
| Preference is given to 1.0L to 1.6L | 0.7 |
| Preference Ix25 | 0.5 |
Recommending the obtained recommendation label to a service person by the recommending device, so that the service person formulates a user grouping rule based on the experience label and the recommendation label, then acquires related user data from the client data storage module based on the formulated user grouping rule, and then obtains a target user group based on the acquired related user data, thereby completing user grouping operation and providing a target service for the target user group. For example, the user can obtain reference information of a certain brand of automobile, and the user can be helped to make an order vehicle plan.
Furthermore, in the embodiment of the invention, the user grouping operation can be fully utilized, and the user grouping rule formulated in the user grouping operation process is reused, so that the aim of updating the reference knowledge graph is fulfilled. Accordingly, step 301 may comprise: the designated tag is periodically obtained based on a predetermined user clustering rule. Step 302 may include: the reference tag information is periodically acquired. For example, the recommendation device obtains the designated tag every other week based on the user clustering rule, and obtains the reference tag information every other week.
Because the user clustering rule is formulated based on the experience label and the recommendation label, the label is obtained based on the user clustering rule, and then the entity corresponding to the label and the reference label information for indicating the user behavior in the reference knowledge map is determined, then establishing an incidence relation for corresponding entities in the reference knowledge graph according to the user behavior indicated by the reference label information and labels obtained based on the user clustering rules, the reference knowledge graph can be further updated and optimized, and finally, the recommended labels determined based on the reference knowledge graph are continuously updated and optimized, that is, the labels with the co-occurrence relationships predicted in table 2 are continuously updated and optimized, the continuous updating and optimizing process improves the efficiency and the accuracy of business personnel for formulating the user grouping rule and improves the efficiency and the accuracy of determining the target user group.
It should be noted that the order of the steps of the tag recommendation method provided in the embodiment of the present invention may be appropriately adjusted. The steps can be increased or decreased according to the circumstances, and any method that can be easily conceived by those skilled in the art within the technical scope of the present disclosure is covered by the protection scope of the present disclosure, and thus, the detailed description thereof is omitted.
In summary, the tag recommendation method provided in the embodiments of the present invention can obtain reference tag information, determine target entities in an initial knowledge graph corresponding to each reference tag of n (n is greater than or equal to 2) reference tags and a pre-obtained designated tag, obtain m (m is greater than or equal to 2) target entities, establish association relationships for the m target entities according to user behaviors and the designated tags indicated by the reference tag information, obtain a reference knowledge graph, and then determine recommendation tags based on the reference knowledge graph. The reference label information is used for indicating user behaviors generated by a user aiming at target services, the reference label information comprises information of n reference labels, the designated labels are obtained based on a preset user clustering rule, and the reference knowledge graph is used for reflecting the incidence relation of m target entities. The method improves the knowledge graph by combining with the user behavior, and recommends the labels for the service personnel based on the improved knowledge graph.
Fig. 4-1 is a schematic structural diagram of a tag recommendation apparatus 400 according to an embodiment of the present invention, where the tag recommendation apparatus 400 may be used in the recommendation device 01 shown in fig. 1, as shown in fig. 4-1, the tag recommendation apparatus 400 includes:
a first obtaining module 410, configured to implement step 302 in the foregoing embodiment.
A first determining module 420, configured to implement step 303 in the foregoing embodiment.
A building module 430, configured to implement step 304 in the foregoing embodiments.
A second determining module 440, configured to implement step 305 in the foregoing embodiment.
Optionally, the reference tag information includes co-occurrence relationship between reference tags obtained by user behavior.
Further, as shown in fig. 4-2, the tag recommendation apparatus 400 may further include:
a second obtaining module 450, configured to implement step 301 in the foregoing embodiment.
As shown in fig. 4-3, the second determining module 440 includes:
a determination submodule 441 is used to implement step 3051 in the above embodiment.
The prediction sub-module 442 is configured to implement the step 3052 in the above embodiment.
An obtaining sub-module 443, configured to implement step 3053 in the foregoing embodiment.
And a query submodule 444 for implementing step 3054 in the above embodiment.
And the processing sub-module 445 is configured to implement step 3055 in the foregoing embodiment.
Optionally, the second obtaining module 450 in fig. 4-2 is configured to implement step 3011 and step 3012 in the foregoing embodiment.
Optionally, the first tag list includes a sorting value corresponding to the tag, and the processing sub-module 445 in fig. 4-3 is configured to implement the step 3055a and the step 3055b in the foregoing embodiment.
Optionally, the first tag list includes an ordering value corresponding to the tag, and the processing sub-module 445 is configured to implement step 3055c and step 3055d in the foregoing embodiment.
Optionally, the second obtaining module 450 in fig. 4-2 is configured to:
the designated tag is periodically obtained based on a predetermined user clustering rule.
The first obtaining module 410 in fig. 4-2, configured to:
the reference tag information is periodically acquired.
The first determining module 420 in fig. 4-2 is configured to implement step 3031 and step 3032 in the foregoing embodiment.
Optionally, the expected tag information includes a tag name of the specified tag, a determination manner of the user clustering rule, a tag statistical period, and a tag entity of the specified tag.
Optionally, the determining sub-module 441 in fig. 4-3 is configured to:
and determining the first vector and the second vector by adopting a knowledge base mode.
Other reference meanings in FIG. 4-2 can be referred to in FIG. 4-1.
In summary, the tag recommendation device provided in the embodiment of the present invention can obtain reference tag information, determine target entities in an initial knowledge graph corresponding to each reference tag of n (n is greater than or equal to 2) reference tags and a pre-obtained designated tag, obtain m (m is greater than or equal to 2) target entities, establish association relationships for the m target entities according to user behaviors and the designated tags indicated by the reference tag information, obtain a reference knowledge graph, and then determine recommendation tags based on the reference knowledge graph. The reference label information is used for indicating user behaviors generated by a user aiming at target services, the reference label information comprises information of n reference labels, the designated labels are obtained based on a preset user clustering rule, and the reference knowledge graph is used for reflecting the incidence relation of m target entities. The device improves the knowledge graph by combining with the user behavior, and recommends the labels for service personnel based on the improved knowledge graph.
It can be clearly understood by those skilled in the art that, for convenience and brevity of description, the specific working processes of the above-described apparatuses and modules may refer to the corresponding processes in the foregoing method embodiments, and are not described herein again.
In the above embodiments, the implementation may be wholly or partially realized by software, hardware, firmware, or any combination thereof. When implemented in software, may be implemented in whole or in part in the form of a computer program product comprising one or more computer instructions. When loaded and executed on a computer, cause the processes or functions described in accordance with the embodiments of the invention to occur, in whole or in part. The computer may be a general purpose computer, a network of computers, or other programmable device. The computer instructions may be stored in a computer readable storage medium or transmitted from one computer readable storage medium to another, for example, the computer instructions may be transmitted from one website, computer, server, or data center to another website, computer, server, or data center by wire (e.g., coaxial cable, fiber optic, digital subscriber line) or wirelessly (e.g., infrared, wireless, microwave, etc.). The computer-readable storage medium can be any available medium that can be accessed by a computer or a data storage device including one or more available media integrated servers, data centers, and the like. The usable medium may be a magnetic medium (e.g., floppy disk, hard disk, magnetic tape), an optical medium, or a semiconductor medium (e.g., solid state disk), among others.
It should be noted that: in the tag recommendation apparatus provided in the above embodiment, when recommending a tag, only the division of the functional modules is illustrated, and in practical applications, the function distribution may be completed by different functional modules according to needs, that is, the internal structure of the device is divided into different functional modules to complete all or part of the functions described above. In addition, the tag recommendation device and the tag recommendation method provided by the embodiments belong to the same concept, and specific implementation processes thereof are described in the method embodiments in detail and are not described herein again.
It will be understood by those skilled in the art that all or part of the steps for implementing the above embodiments may be implemented by hardware, or may be implemented by a program instructing relevant hardware, where the program may be stored in a computer-readable storage medium, and the above-mentioned storage medium may be a read-only memory, a magnetic disk or an optical disk, etc.
The above description is only exemplary of the present application and should not be taken as limiting, as any modification, equivalent replacement, or improvement made within the spirit and principle of the present application should be included in the protection scope of the present application.
Claims (20)
1. A tag recommendation method, the method comprising:
acquiring reference label information, wherein the reference label information is used for indicating user behaviors generated by a user aiming at a target service, the reference label information comprises information of n reference labels, n is more than or equal to 2, and the reference label information comprises a co-occurrence relation between the reference labels obtained by the user behaviors;
determining target entities corresponding to each reference label in the n reference labels and a pre-acquired designated label in the initial knowledge graph to obtain m target entities, wherein m is larger than or equal to 2, and the designated label is acquired based on a preset user clustering rule;
establishing association relations for the m target entities according to the user behaviors indicated by the reference label information and the designated labels to obtain a reference knowledge graph, wherein the reference knowledge graph is used for reflecting the association relations of the m target entities;
determining a recommended label based on the reference knowledge-graph;
prior to the obtaining reference tag information, the method further comprises:
acquiring the designated label based on the predetermined user grouping rule, wherein the user grouping rule is determined based on the target service;
the determining a recommended label based on the reference knowledge-graph comprises:
determining a first vector of a first target entity and a second vector of an incidence relation related to the first target entity in the reference knowledge-graph, wherein the first target entity is any one of the m target entities;
predicting an entity having a co-occurrence relationship with the first target entity by adopting a scoring function according to the first vector and the second vector to obtain expected entity information, wherein the expected entity information comprises the first target entity, the entity having the co-occurrence relationship with the first target entity and a scoring result corresponding to the first target entity;
obtaining expected label information corresponding to the specified label;
querying a tag corresponding to a target entity in the expected entity information from the expected tag information, and obtaining a first tag list based on the expected entity information;
and performing data redundancy processing on the first tag list to obtain a recommendation list, wherein the recommendation list comprises the recommendation tags.
2. The method of claim 1, wherein obtaining the designated label based on the predetermined user grouping rule comprises:
storing the user grouping rule in a text form;
and extracting the specified label from the user clustering rule in a text form.
3. The method of claim 1, wherein the first tag list comprises a ranking value corresponding to a tag,
the performing data redundancy processing on the first tag list to obtain a recommendation list includes:
de-duplicating and reserving one of a plurality of identical tags which repeatedly appear in the first tag list;
and taking the sum of the ranking values corresponding to each label in the same labels as the preserved ranking value of the corresponding label.
4. The method of claim 1, wherein the first tag list comprises a ranking value corresponding to a tag,
the performing data redundancy processing on the first tag list to obtain a recommendation list includes:
de-duplicating and reserving one of a plurality of identical tags which repeatedly appear in the first tag list;
and carrying out weighted summation on the ranking value corresponding to each label in the same labels, and taking the result of the weighted summation as the preserved ranking value corresponding to the label.
5. The method of claim 1, wherein obtaining the designated label based on the predetermined user grouping rule comprises:
and periodically acquiring the designated label based on the preset user grouping rule.
6. The method of claim 5, wherein the obtaining reference tag information comprises:
and periodically acquiring the reference label information.
7. The method of claim 1, wherein the determining the target entities in the initial knowledge-graph corresponding to each of the n reference tags and a pre-obtained designated tag, resulting in m target entities comprises:
filtering preset entities in the initial knowledge graph by adopting a word frequency TF reverse file frequency IDF statistical mode;
determining the m target entities from the filtered entities.
8. The method of claim 1,
the expected tag information comprises the tag name of the specified tag, the determination mode of the user grouping rule, the tag statistical period and the tag entity of the specified tag.
9. The method of claim 3,
the determining a first vector of a first target entity and a second vector of an associative relationship associated with the first target entity in the reference knowledge-graph comprises:
and determining the first vector and the second vector by adopting a knowledge base mode.
10. A tag recommendation apparatus, the apparatus comprising:
the first acquisition module is used for acquiring reference label information, wherein the reference label information is used for indicating user behaviors generated by a user aiming at a target service, the reference label information comprises information of n reference labels, n is more than or equal to 2, and the reference label information comprises a co-occurrence relation between the reference labels obtained by the user behaviors;
a first determining module, configured to determine a target entity corresponding to each of the n reference tags and a pre-obtained designated tag in an initial knowledge graph, to obtain m target entities, where m is greater than or equal to 2, and the designated tag is obtained based on a predetermined user clustering rule;
the establishing module is used for establishing association relations for the m target entities according to the user behaviors indicated by the reference label information and the designated labels to obtain a reference knowledge graph, and the reference knowledge graph is used for reflecting the association relations of the m target entities;
a second determination module to determine a recommended label based on the reference knowledge-graph;
the device further comprises:
a second obtaining module, configured to obtain the designated tag based on the predetermined user grouping rule, where the user grouping rule is determined based on the target service;
the second determining module includes:
a determining sub-module, configured to determine a first vector of a first target entity and a second vector of an association relation related to the first target entity in the reference knowledge-graph, where the first target entity is any one of the m target entities;
the prediction sub-module is used for predicting an entity which has a co-occurrence relationship with the first target entity by adopting a scoring function according to the first vector and the second vector to obtain expected entity information, wherein the expected entity information comprises the first target entity, the entity which has the co-occurrence relationship with the first target entity and a scoring result corresponding to the first target entity;
the obtaining submodule is used for obtaining expected tag information corresponding to the specified tag;
the query submodule is used for querying a tag corresponding to a target entity in the expected entity information from the expected tag information and obtaining a first tag list based on the expected entity information;
and the processing submodule is used for performing data redundancy processing on the first tag list to obtain a recommendation list, and the recommendation list comprises the recommendation tags.
11. The apparatus of claim 10, wherein the second obtaining module is configured to:
storing the user grouping rule in a text form;
and extracting the specified label from the user clustering rule in a text form.
12. The apparatus of claim 10, wherein the first tag list comprises a ranking value corresponding to a tag,
the processing submodule is used for:
de-duplicating and reserving one of a plurality of identical tags which repeatedly appear in the first tag list;
and taking the sum of the ranking values corresponding to each label in the same labels as the preserved ranking value of the corresponding label.
13. The apparatus of claim 10, wherein the first tag list comprises a ranking value corresponding to a tag,
the processing submodule is used for:
de-duplicating and reserving one of a plurality of identical tags which repeatedly appear in the first tag list;
and carrying out weighted summation on the ranking value corresponding to each label in the same labels, and taking the result of the weighted summation as the preserved ranking value corresponding to the label.
14. The apparatus of claim 10, wherein the second obtaining module is configured to:
and periodically acquiring the designated label based on the preset user grouping rule.
15. The apparatus of claim 14, wherein the first obtaining module is configured to:
and periodically acquiring the reference label information.
16. The apparatus of claim 10, wherein the first determining module is configured to:
filtering preset entities in the initial knowledge graph by adopting a word frequency TF reverse file frequency IDF statistical mode;
determining the m target entities from the filtered entities.
17. The apparatus of claim 10,
the expected tag information comprises the tag name of the specified tag, the determination mode of the user grouping rule, the tag statistical period and the tag entity of the specified tag.
18. The apparatus of claim 10,
the determination submodule is configured to:
and determining the first vector and the second vector by adopting a knowledge base mode.
19. A tag recommendation device comprising a memory, a processor and a program stored on said memory and executable on said processor, said processor executing said program to perform the tag recommendation method of any of claims 1 to 9.
20. A computer-readable storage medium having stored therein instructions which, when run on a computer, cause the computer to perform the tag recommendation method of any one of claims 1 to 9.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710429233.7A CN109033101B (en) | 2017-06-08 | 2017-06-08 | Label recommendation method and device |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710429233.7A CN109033101B (en) | 2017-06-08 | 2017-06-08 | Label recommendation method and device |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN109033101A CN109033101A (en) | 2018-12-18 |
| CN109033101B true CN109033101B (en) | 2021-08-20 |
Family
ID=64629622
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201710429233.7A Active CN109033101B (en) | 2017-06-08 | 2017-06-08 | Label recommendation method and device |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109033101B (en) |
Families Citing this family (13)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111522886B (en) * | 2019-01-17 | 2023-05-09 | 中国移动通信有限公司研究院 | Information recommendation method, terminal and storage medium |
| CN111523010B (en) * | 2019-02-03 | 2023-04-28 | 阿里巴巴集团控股有限公司 | Recommendation method, recommendation device, terminal equipment and computer storage medium |
| CN110210892B (en) * | 2019-05-05 | 2023-05-30 | 平安科技（深圳）有限公司 | Product recommendation method, device and readable storage medium |
| CN110321482B (en) * | 2019-06-11 | 2023-04-18 | 创新先进技术有限公司 | Information recommendation method, device and equipment |
| CN110674307A (en) * | 2019-08-21 | 2020-01-10 | 北京邮电大学 | Knowledge deduction method and system for knowledge center network |
| CN111091454A (en) * | 2019-11-05 | 2020-05-01 | 新华智云科技有限公司 | Financial public opinion recommendation method based on knowledge graph |
| CN111078972B (en) * | 2019-11-29 | 2023-06-16 | 支付宝(杭州)信息技术有限公司 | Questioning behavior data acquisition method, questioning behavior data acquisition device and server |
| CN112084761B (en) * | 2020-09-02 | 2024-09-06 | 董萍 | Hydraulic engineering information management method and device |
| CN112632405B (en) * | 2020-12-31 | 2024-05-10 | 数字广东网络建设有限公司 | Recommendation method, recommendation device, recommendation equipment and storage medium |
| CN112765398A (en) * | 2021-01-04 | 2021-05-07 | 珠海格力电器股份有限公司 | Information recommendation method and device and storage medium |
| CN114417007A (en) * | 2021-12-21 | 2022-04-29 | 建信金融科技有限责任公司 | Recommendation method and related equipment for financial products |
| CN114996507B (en) * | 2022-06-10 | 2024-08-06 | 北京达佳互联信息技术有限公司 | Video recommendation method and device |
| CN117579456A (en) * | 2023-10-18 | 2024-02-20 | 中移互联网有限公司 | Service message sending method and device, electronic equipment and storage medium |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20100185509A1 (en) * | 2009-01-21 | 2010-07-22 | Yahoo! Inc. | Interest-based ranking system for targeted marketing |
| CN103177384A (en) * | 2013-04-16 | 2013-06-26 | 高鹏 | Network advertisement putting method based on user interest spectrum |
| CN103577549A (en) * | 2013-10-16 | 2014-02-12 | 复旦大学 | Crowd portrayal system and method based on microblog label |
| CN106294744A (en) * | 2016-08-11 | 2017-01-04 | 上海动云信息科技有限公司 | Interest recognition methods and system |
-
2017
- 2017-06-08 CN CN201710429233.7A patent/CN109033101B/en active Active
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20100185509A1 (en) * | 2009-01-21 | 2010-07-22 | Yahoo! Inc. | Interest-based ranking system for targeted marketing |
| CN103177384A (en) * | 2013-04-16 | 2013-06-26 | 高鹏 | Network advertisement putting method based on user interest spectrum |
| CN103577549A (en) * | 2013-10-16 | 2014-02-12 | 复旦大学 | Crowd portrayal system and method based on microblog label |
| CN106294744A (en) * | 2016-08-11 | 2017-01-04 | 上海动云信息科技有限公司 | Interest recognition methods and system |
Also Published As
| Publication number | Publication date |
|---|---|
| CN109033101A (en) | 2018-12-18 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109033101B (en) | Label recommendation method and device | |
| US10853360B2 (en) | Searchable index | |
| CN108804641B (en) | Text similarity calculation method, device, equipment and storage medium | |
| US8589429B1 (en) | System and method for providing query recommendations based on search activity of a user base | |
| US9934293B2 (en) | Generating search results | |
| US10838985B2 (en) | Item recommendation method, device, and system | |
| CA3059929C (en) | Text searching method, apparatus, and non-transitory computer-readable storage medium | |
| CN108717407B (en) | Entity vector determination method and device, and information retrieval method and device | |
| CN106844407B (en) | Tag network generation method and system based on data set correlation | |
| US20170371965A1 (en) | Method and system for dynamically personalizing profiles in a social network | |
| CN109241243B (en) | Candidate document sorting method and device | |
| CN110795613B (en) | Commodity searching method, device and system and electronic equipment | |
| US9552415B2 (en) | Category classification processing device and method | |
| CN111444304A (en) | Search ranking method and device | |
| CN112100396A (en) | Data processing method and device | |
| CN111078997A (en) | Information recommendation method and device | |
| CN111160699A (en) | Expert recommendation method and system | |
| CN116610853A (en) | Search recommendation method, search recommendation system, computer device, and storage medium | |
| US20150206220A1 (en) | Recommendation Strategy Portfolios | |
| CN107798045A (en) | User towards middle-size and small-size website, which accesses, is intended to acquisition methods and system | |
| CN112288510A (en) | Article recommendation method, device, equipment and storage medium | |
| CN108984582B (en) | Query request processing method | |
| CA2485814A1 (en) | Method and apparatus for range processing in an n-dimensional space | |
| CN113763084A (en) | Product recommendation processing method, device, equipment and storage medium | |
| CN115795023B (en) | Document recommendation method, device, equipment and storage medium |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| TA01 | Transfer of patent application right | ||
| TA01 | Transfer of patent application right |
Effective date of registration: 20200201 Address after: 518129 Bantian HUAWEI headquarters office building, Longgang District, Guangdong, Shenzhen Applicant after: HUAWEI TECHNOLOGIES Co.,Ltd. Address before: 210000 HUAWEI Nanjing base, 101 software Avenue, Yuhuatai District, Jiangsu, Nanjing Applicant before: Huawei Technologies Co.,Ltd. |
|
| GR01 | Patent grant | ||
| GR01 | Patent grant |